IRL SF Cleaning Service

IRL SF Cleaning Service provides a ready-to-use MCP server that books cleaning services for San Francisco. You can run it locally or deploy to a remote host (Vercel) for outside access, and connect to it with an MCP client to issue booking commands. The server can email partners via Resend and supports testing and integration through clearly defined steps.

How to use

You will interact with the server through an MCP client to invoke the available tools. The primary action is to request a cleaning service for a San Francisco address. If your input includes SF-related address data, the server will process the request and, upon successful validation, trigger a booking workflow and notification emails. If the address is outside SF, you will receive a polite rejection indicating that SF is currently the service area.

The server exposes a single tool named request_cleaning. Use your MCP client to call this tool with the customer name, contact phone, and a service address. The client should handle the response by presenting either a confirmation message (for SF addresses) or a clear rejection (for non-SF addresses). You can test the remote server URL after deployment to verify that the tool is available and returns the expected results.

How to install

Prerequisites: you need Node.js and npm installed on your machine. Ensure you have a working internet connection for dependency installation and remote deployment.

# 1. Install dependencies for the MCP server
npm install

2. Set up email sending credentials using Resend and prepare environment variables. You will configure a Resend API key, a From email, and partner emails in the environment file.

RESEND_API_KEY=re_xxxxxxxxxxxxxxxxxx
FROM_EMAIL=gwen@irl-concierge.com
PARTNER_EMAILS=partner1@example.com,partner2@example.com

3. Create or update your environment file with the Resend credentials and the sender information. You can use onboarding for testing if the domain is not yet verified.

RESEND_API_KEY=re_xxxxxxxxxxxxxxxxxx
FROM_EMAIL=gwen@irl-concierge.com
PARTNER_EMAILS=partner1@example.com,partner2@example.com

4. Configure Claude Desktop to point to your local MCP server. Update the configuration with the path to the MCP entry file.

{
  "mcpServers": {
    "sf-cleaning": {
      "command": "node",
      "args": ["/absolute/path/to/mcp-sf-cleaning/index.js"]
    }
  }
}

Additional deployment and testing options

Remote deployment through Vercel lets you expose the MCP server to the internet. You will install the Vercel CLI, provide environment variables, deploy, and then connect your Claude Desktop to the resulting URL.

Steps to deploy on Vercel and use the remote MCP URL include installing the Vercel CLI, setting environment variables, running the deployment, and then testing the remote endpoint.

After deployment, you should receive a remote MCP URL similar to https://mcp-sf-cleaning.vercel.app/api/mcp and add this as a custom connector in Claude Desktop.

For testing the remote server directly, you can submit a request to the MCP URL using your preferred HTTP client to list tools or call the request_cleaning tool with the required parameters (name, phone, and SF address).

Examples and testing notes

You can verify the server behavior by performing a simple tool inquiry and a booking request against the remote endpoint. The server will respond with a confirmation message for SF addresses or a rejection for non-SF addresses.

Available tools

request_cleaning

Book a cleaning service for an address in San Francisco. Requires customer name, phone, and SF address. Returns a confirmation if SF is covered or a polite rejection otherwise.